CodeBunny isn't just a linter. It uses a multi-agent architecture to understand your repository's context, catch architectural risks, and speed up reviews by 10x.
I've analyzed the changes in src/auth/jwt_handler.py against the repository context.
Refactoring token validation to support the new SSO provider. This aligns with the migration plan outlined in RFC-002.
High Risk: Modifies core authentication logic. This change bypasses the legacy cache layer. Verify load testing on the auth service before merge.
Powered by Next-Gen Tech
Most tools summarize blindly. CodeBunny uses a sophisticated Map-Reduce architecture to handle massive PRs without losing context.
The system receives the webhook. It fetches the raw Diff and the full Repository File Tree to build a context graph.
The Diff is split. Mistral Small agents run in parallel, analyzing individual files for technical accuracy.
Gemini 1.5 Pro (The Strategist) ingests all file summaries + the file tree to synthesize the "Three-Pillar Analysis."
A structured comment is posted. It highlights risks, architectural impact, and missed patterns.
Won't even be requiring your put your tokens in the first place. CodeBunny's architecture is designed to run with zero access to your codebase.
FastAPI backend running on Docker containers ensures typical reviews land in < 30 seconds.
Context is king. We analyze your file structure to know if a change in utils.py breaks core/billing.
Don't trust the cloud? Run CodeBunny on your own infrastructure. Docker-ready, configurable, and free for the community.
View Source| Feature | CodeBunny | Standard LLM Bots |
|---|---|---|
| Context Window | Infinite (Map-Reduce) | Limited (Truncates files) |
| Repo Structure Awareness | Yes (File Tree analysis) | No (Diff only) |
| Reasoning Engine | Multi-Model (Mistral + Gemini) | Single generic model |
| Cost | Free / Open Source | $$$ per seat |
Join hundreds of developers who trust CodeBunny to review their PRs.